Model save
Browse files
README.md
ADDED
|
@@ -0,0 +1,80 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 3 |
+
library_name: peft
|
| 4 |
+
license: apache-2.0
|
| 5 |
+
tags:
|
| 6 |
+
- trl
|
| 7 |
+
- dpo
|
| 8 |
+
- generated_from_trainer
|
| 9 |
+
model-index:
|
| 10 |
+
- name: zephyr-dpo-qlora-uf-ours-5e-6-epoch1
|
| 11 |
+
results: []
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 15 |
+
should probably proofread and complete it, then remove this comment. -->
|
| 16 |
+
|
| 17 |
+
# zephyr-dpo-qlora-uf-ours-5e-6-epoch1
|
| 18 |
+
|
| 19 |
+
This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on the None dataset.
|
| 20 |
+
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.8935
|
| 22 |
+
- Rewards/chosen: -4.8886
|
| 23 |
+
- Rewards/rejected: -5.8224
|
| 24 |
+
- Rewards/accuracies: 0.6470
|
| 25 |
+
- Rewards/margins: 0.9339
|
| 26 |
+
- Rewards/margins Max: 4.5089
|
| 27 |
+
- Rewards/margins Min: -2.8033
|
| 28 |
+
- Rewards/margins Std: 2.4986
|
| 29 |
+
- Logps/rejected: -840.8231
|
| 30 |
+
- Logps/chosen: -773.4503
|
| 31 |
+
- Logits/rejected: -1.3417
|
| 32 |
+
- Logits/chosen: -1.4023
|
| 33 |
+
|
| 34 |
+
## Model description
|
| 35 |
+
|
| 36 |
+
More information needed
|
| 37 |
+
|
| 38 |
+
## Intended uses & limitations
|
| 39 |
+
|
| 40 |
+
More information needed
|
| 41 |
+
|
| 42 |
+
## Training and evaluation data
|
| 43 |
+
|
| 44 |
+
More information needed
|
| 45 |
+
|
| 46 |
+
## Training procedure
|
| 47 |
+
|
| 48 |
+
### Training hyperparameters
|
| 49 |
+
|
| 50 |
+
The following hyperparameters were used during training:
|
| 51 |
+
- learning_rate: 5e-06
|
| 52 |
+
- train_batch_size: 4
|
| 53 |
+
- eval_batch_size: 8
|
| 54 |
+
- seed: 42
|
| 55 |
+
- distributed_type: multi-GPU
|
| 56 |
+
- num_devices: 2
|
| 57 |
+
- gradient_accumulation_steps: 2
|
| 58 |
+
- total_train_batch_size: 16
|
| 59 |
+
- total_eval_batch_size: 16
|
| 60 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 61 |
+
- lr_scheduler_type: cosine
|
| 62 |
+
- lr_scheduler_warmup_ratio: 0.1
|
| 63 |
+
- num_epochs: 1
|
| 64 |
+
|
| 65 |
+
### Training results
|
| 66 |
+
|
| 67 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Rewards/margins Max | Rewards/margins Min | Rewards/margins Std | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 68 |
+
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:-------------------:|:-------------------:|:-------------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 69 |
+
| 0.4419 | 0.28 | 100 | 0.6650 | -0.4529 | -0.5977 | 0.6240 | 0.1449 | 0.8619 | -0.5418 | 0.4692 | -318.3518 | -329.8811 | -2.5449 | -2.5719 |
|
| 70 |
+
| 0.1872 | 0.56 | 200 | 0.7859 | -2.8392 | -3.5400 | 0.6630 | 0.7008 | 3.4374 | -2.1827 | 1.9158 | -612.5828 | -568.5178 | -1.4159 | -1.4771 |
|
| 71 |
+
| 0.1102 | 0.85 | 300 | 0.8935 | -4.8886 | -5.8224 | 0.6470 | 0.9339 | 4.5089 | -2.8033 | 2.4986 | -840.8231 | -773.4503 | -1.3417 | -1.4023 |
|
| 72 |
+
|
| 73 |
+
|
| 74 |
+
### Framework versions
|
| 75 |
+
|
| 76 |
+
- PEFT 0.7.1
|
| 77 |
+
- Transformers 4.39.0.dev0
|
| 78 |
+
- Pytorch 2.1.2+cu121
|
| 79 |
+
- Datasets 2.14.6
|
| 80 |
+
- Tokenizers 0.15.2
|
adapter_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 671150064
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8929f13332d1c0d3d17af32b08282af151e831d5361900cb1c75c6d84d6d704a
|
| 3 |
size 671150064
|
all_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.3073710792501208,
|
| 4 |
+
"train_runtime": 4217.5244,
|
| 5 |
+
"train_samples": 5678,
|
| 6 |
+
"train_samples_per_second": 1.346,
|
| 7 |
+
"train_steps_per_second": 0.084
|
| 8 |
+
}
|
runs/Jul29_11-02-20_notebook-deployment-48-7d9b6c99-khd85/events.out.tfevents.1722251041.notebook-deployment-48-7d9b6c99-khd85.3446415.0
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dbe9bc7605d1642b494dffc8cf3ff6bdc3dac1e5e352d0c497a23ba0c4ac1d5c
|
| 3 |
+
size 39981
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.3073710792501208,
|
| 4 |
+
"train_runtime": 4217.5244,
|
| 5 |
+
"train_samples": 5678,
|
| 6 |
+
"train_samples_per_second": 1.346,
|
| 7 |
+
"train_steps_per_second": 0.084
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,735 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 355,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.0,
|
| 13 |
+
"grad_norm": 1.6023182644205256,
|
| 14 |
+
"learning_rate": 1.3888888888888888e-07,
|
| 15 |
+
"logits/chosen": -2.861618995666504,
|
| 16 |
+
"logits/rejected": -2.8205904960632324,
|
| 17 |
+
"logps/chosen": -271.06011962890625,
|
| 18 |
+
"logps/rejected": -211.1704559326172,
|
| 19 |
+
"loss": 0.6931,
|
| 20 |
+
"rewards/accuracies": 0.0,
|
| 21 |
+
"rewards/chosen": 0.0,
|
| 22 |
+
"rewards/margins": 0.0,
|
| 23 |
+
"rewards/margins_max": 0.0,
|
| 24 |
+
"rewards/margins_min": 0.0,
|
| 25 |
+
"rewards/margins_std": 0.0,
|
| 26 |
+
"rewards/rejected": 0.0,
|
| 27 |
+
"step": 1
|
| 28 |
+
},
|
| 29 |
+
{
|
| 30 |
+
"epoch": 0.03,
|
| 31 |
+
"grad_norm": 9.801703666985174,
|
| 32 |
+
"learning_rate": 1.3888888888888892e-06,
|
| 33 |
+
"logits/chosen": -2.833193778991699,
|
| 34 |
+
"logits/rejected": -2.7905588150024414,
|
| 35 |
+
"logps/chosen": -324.9515380859375,
|
| 36 |
+
"logps/rejected": -274.97601318359375,
|
| 37 |
+
"loss": 0.6927,
|
| 38 |
+
"rewards/accuracies": 0.5,
|
| 39 |
+
"rewards/chosen": 0.0009881681762635708,
|
| 40 |
+
"rewards/margins": 0.001091944519430399,
|
| 41 |
+
"rewards/margins_max": 0.005423115566372871,
|
| 42 |
+
"rewards/margins_min": -0.0020756451413035393,
|
| 43 |
+
"rewards/margins_std": 0.0033777032513171434,
|
| 44 |
+
"rewards/rejected": -0.00010377627768320963,
|
| 45 |
+
"step": 10
|
| 46 |
+
},
|
| 47 |
+
{
|
| 48 |
+
"epoch": 0.06,
|
| 49 |
+
"grad_norm": 1.8295784413634215,
|
| 50 |
+
"learning_rate": 2.7777777777777783e-06,
|
| 51 |
+
"logits/chosen": -2.7249505519866943,
|
| 52 |
+
"logits/rejected": -2.7066354751586914,
|
| 53 |
+
"logps/chosen": -293.10626220703125,
|
| 54 |
+
"logps/rejected": -215.7418670654297,
|
| 55 |
+
"loss": 0.6894,
|
| 56 |
+
"rewards/accuracies": 0.800000011920929,
|
| 57 |
+
"rewards/chosen": 0.007196377031505108,
|
| 58 |
+
"rewards/margins": 0.0068252841010689735,
|
| 59 |
+
"rewards/margins_max": 0.015210196375846863,
|
| 60 |
+
"rewards/margins_min": -0.0003913335967808962,
|
| 61 |
+
"rewards/margins_std": 0.006930469069629908,
|
| 62 |
+
"rewards/rejected": 0.0003710930759552866,
|
| 63 |
+
"step": 20
|
| 64 |
+
},
|
| 65 |
+
{
|
| 66 |
+
"epoch": 0.08,
|
| 67 |
+
"grad_norm": 2.0991251258429813,
|
| 68 |
+
"learning_rate": 4.166666666666667e-06,
|
| 69 |
+
"logits/chosen": -2.816267490386963,
|
| 70 |
+
"logits/rejected": -2.7468106746673584,
|
| 71 |
+
"logps/chosen": -300.4118347167969,
|
| 72 |
+
"logps/rejected": -232.51480102539062,
|
| 73 |
+
"loss": 0.6751,
|
| 74 |
+
"rewards/accuracies": 0.887499988079071,
|
| 75 |
+
"rewards/chosen": 0.03294036164879799,
|
| 76 |
+
"rewards/margins": 0.03294597938656807,
|
| 77 |
+
"rewards/margins_max": 0.06541319936513901,
|
| 78 |
+
"rewards/margins_min": 0.006184346973896027,
|
| 79 |
+
"rewards/margins_std": 0.02764921821653843,
|
| 80 |
+
"rewards/rejected": -5.6155026868509594e-06,
|
| 81 |
+
"step": 30
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"epoch": 0.11,
|
| 85 |
+
"grad_norm": 1.7320040744134495,
|
| 86 |
+
"learning_rate": 4.998060489154965e-06,
|
| 87 |
+
"logits/chosen": -2.8301820755004883,
|
| 88 |
+
"logits/rejected": -2.751152753829956,
|
| 89 |
+
"logps/chosen": -271.0854187011719,
|
| 90 |
+
"logps/rejected": -225.5413360595703,
|
| 91 |
+
"loss": 0.6584,
|
| 92 |
+
"rewards/accuracies": 0.875,
|
| 93 |
+
"rewards/chosen": 0.06399233639240265,
|
| 94 |
+
"rewards/margins": 0.0620587058365345,
|
| 95 |
+
"rewards/margins_max": 0.128224715590477,
|
| 96 |
+
"rewards/margins_min": 0.005723648704588413,
|
| 97 |
+
"rewards/margins_std": 0.05614202097058296,
|
| 98 |
+
"rewards/rejected": 0.0019336309051141143,
|
| 99 |
+
"step": 40
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.14,
|
| 103 |
+
"grad_norm": 1.9286136496428266,
|
| 104 |
+
"learning_rate": 4.976275538042932e-06,
|
| 105 |
+
"logits/chosen": -2.7784149646759033,
|
| 106 |
+
"logits/rejected": -2.7074830532073975,
|
| 107 |
+
"logps/chosen": -263.76910400390625,
|
| 108 |
+
"logps/rejected": -236.1763458251953,
|
| 109 |
+
"loss": 0.6279,
|
| 110 |
+
"rewards/accuracies": 0.9375,
|
| 111 |
+
"rewards/chosen": 0.1180100291967392,
|
| 112 |
+
"rewards/margins": 0.14100034534931183,
|
| 113 |
+
"rewards/margins_max": 0.2934209704399109,
|
| 114 |
+
"rewards/margins_min": 0.029913101345300674,
|
| 115 |
+
"rewards/margins_std": 0.12193763256072998,
|
| 116 |
+
"rewards/rejected": -0.022990306839346886,
|
| 117 |
+
"step": 50
|
| 118 |
+
},
|
| 119 |
+
{
|
| 120 |
+
"epoch": 0.17,
|
| 121 |
+
"grad_norm": 2.19542131799087,
|
| 122 |
+
"learning_rate": 4.93049306999712e-06,
|
| 123 |
+
"logits/chosen": -2.6885108947753906,
|
| 124 |
+
"logits/rejected": -2.653852939605713,
|
| 125 |
+
"logps/chosen": -303.560791015625,
|
| 126 |
+
"logps/rejected": -276.5751647949219,
|
| 127 |
+
"loss": 0.5911,
|
| 128 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 129 |
+
"rewards/chosen": 0.10078845918178558,
|
| 130 |
+
"rewards/margins": 0.23882508277893066,
|
| 131 |
+
"rewards/margins_max": 0.41835322976112366,
|
| 132 |
+
"rewards/margins_min": 0.06585223972797394,
|
| 133 |
+
"rewards/margins_std": 0.1579587757587433,
|
| 134 |
+
"rewards/rejected": -0.1380366086959839,
|
| 135 |
+
"step": 60
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"epoch": 0.2,
|
| 139 |
+
"grad_norm": 1.8726629315170749,
|
| 140 |
+
"learning_rate": 4.861156761634014e-06,
|
| 141 |
+
"logits/chosen": -2.6910977363586426,
|
| 142 |
+
"logits/rejected": -2.637085437774658,
|
| 143 |
+
"logps/chosen": -308.81890869140625,
|
| 144 |
+
"logps/rejected": -251.6500701904297,
|
| 145 |
+
"loss": 0.5567,
|
| 146 |
+
"rewards/accuracies": 0.9375,
|
| 147 |
+
"rewards/chosen": 0.14042340219020844,
|
| 148 |
+
"rewards/margins": 0.3005082309246063,
|
| 149 |
+
"rewards/margins_max": 0.6100845336914062,
|
| 150 |
+
"rewards/margins_min": 0.049360163509845734,
|
| 151 |
+
"rewards/margins_std": 0.25483688712120056,
|
| 152 |
+
"rewards/rejected": -0.16008484363555908,
|
| 153 |
+
"step": 70
|
| 154 |
+
},
|
| 155 |
+
{
|
| 156 |
+
"epoch": 0.23,
|
| 157 |
+
"grad_norm": 3.2506013460624406,
|
| 158 |
+
"learning_rate": 4.7689385491773934e-06,
|
| 159 |
+
"logits/chosen": -2.6962907314300537,
|
| 160 |
+
"logits/rejected": -2.648644208908081,
|
| 161 |
+
"logps/chosen": -310.2759704589844,
|
| 162 |
+
"logps/rejected": -320.2915344238281,
|
| 163 |
+
"loss": 0.5126,
|
| 164 |
+
"rewards/accuracies": 0.949999988079071,
|
| 165 |
+
"rewards/chosen": 0.1099453717470169,
|
| 166 |
+
"rewards/margins": 0.4490734934806824,
|
| 167 |
+
"rewards/margins_max": 0.8111162185668945,
|
| 168 |
+
"rewards/margins_min": 0.09584008157253265,
|
| 169 |
+
"rewards/margins_std": 0.3334144353866577,
|
| 170 |
+
"rewards/rejected": -0.3391281068325043,
|
| 171 |
+
"step": 80
|
| 172 |
+
},
|
| 173 |
+
{
|
| 174 |
+
"epoch": 0.25,
|
| 175 |
+
"grad_norm": 3.7658324125158935,
|
| 176 |
+
"learning_rate": 4.654732116743193e-06,
|
| 177 |
+
"logits/chosen": -2.5777134895324707,
|
| 178 |
+
"logits/rejected": -2.555579423904419,
|
| 179 |
+
"logps/chosen": -265.8702697753906,
|
| 180 |
+
"logps/rejected": -246.2513427734375,
|
| 181 |
+
"loss": 0.456,
|
| 182 |
+
"rewards/accuracies": 0.987500011920929,
|
| 183 |
+
"rewards/chosen": 0.10124516487121582,
|
| 184 |
+
"rewards/margins": 0.5738865733146667,
|
| 185 |
+
"rewards/margins_max": 0.9613016843795776,
|
| 186 |
+
"rewards/margins_min": 0.21029099822044373,
|
| 187 |
+
"rewards/margins_std": 0.3518252670764923,
|
| 188 |
+
"rewards/rejected": -0.4726414680480957,
|
| 189 |
+
"step": 90
|
| 190 |
+
},
|
| 191 |
+
{
|
| 192 |
+
"epoch": 0.28,
|
| 193 |
+
"grad_norm": 7.900318438715333,
|
| 194 |
+
"learning_rate": 4.5196442356717526e-06,
|
| 195 |
+
"logits/chosen": -2.587555408477783,
|
| 196 |
+
"logits/rejected": -2.5667903423309326,
|
| 197 |
+
"logps/chosen": -279.7422790527344,
|
| 198 |
+
"logps/rejected": -321.7631530761719,
|
| 199 |
+
"loss": 0.4419,
|
| 200 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 201 |
+
"rewards/chosen": 0.01669424958527088,
|
| 202 |
+
"rewards/margins": 0.6124511957168579,
|
| 203 |
+
"rewards/margins_max": 1.1129316091537476,
|
| 204 |
+
"rewards/margins_min": 0.16525402665138245,
|
| 205 |
+
"rewards/margins_std": 0.43622416257858276,
|
| 206 |
+
"rewards/rejected": -0.5957569479942322,
|
| 207 |
+
"step": 100
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"epoch": 0.28,
|
| 211 |
+
"eval_logits/chosen": -2.5718588829040527,
|
| 212 |
+
"eval_logits/rejected": -2.544903039932251,
|
| 213 |
+
"eval_logps/chosen": -329.8811340332031,
|
| 214 |
+
"eval_logps/rejected": -318.3518371582031,
|
| 215 |
+
"eval_loss": 0.6649842262268066,
|
| 216 |
+
"eval_rewards/accuracies": 0.6240000128746033,
|
| 217 |
+
"eval_rewards/chosen": -0.4528772234916687,
|
| 218 |
+
"eval_rewards/margins": 0.14485196769237518,
|
| 219 |
+
"eval_rewards/margins_max": 0.8618865609169006,
|
| 220 |
+
"eval_rewards/margins_min": -0.5417762994766235,
|
| 221 |
+
"eval_rewards/margins_std": 0.46923625469207764,
|
| 222 |
+
"eval_rewards/rejected": -0.5977292656898499,
|
| 223 |
+
"eval_runtime": 428.7091,
|
| 224 |
+
"eval_samples_per_second": 4.665,
|
| 225 |
+
"eval_steps_per_second": 0.292,
|
| 226 |
+
"step": 100
|
| 227 |
+
},
|
| 228 |
+
{
|
| 229 |
+
"epoch": 0.31,
|
| 230 |
+
"grad_norm": 3.5447046302664655,
|
| 231 |
+
"learning_rate": 4.364984038837727e-06,
|
| 232 |
+
"logits/chosen": -2.6017634868621826,
|
| 233 |
+
"logits/rejected": -2.543693780899048,
|
| 234 |
+
"logps/chosen": -371.7081604003906,
|
| 235 |
+
"logps/rejected": -368.0250549316406,
|
| 236 |
+
"loss": 0.3877,
|
| 237 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 238 |
+
"rewards/chosen": 0.028975462540984154,
|
| 239 |
+
"rewards/margins": 0.844887912273407,
|
| 240 |
+
"rewards/margins_max": 1.4032989740371704,
|
| 241 |
+
"rewards/margins_min": 0.22432711720466614,
|
| 242 |
+
"rewards/margins_std": 0.5291948914527893,
|
| 243 |
+
"rewards/rejected": -0.8159124255180359,
|
| 244 |
+
"step": 110
|
| 245 |
+
},
|
| 246 |
+
{
|
| 247 |
+
"epoch": 0.34,
|
| 248 |
+
"grad_norm": 2.8541926547535676,
|
| 249 |
+
"learning_rate": 4.192250333880045e-06,
|
| 250 |
+
"logits/chosen": -2.566699504852295,
|
| 251 |
+
"logits/rejected": -2.523707866668701,
|
| 252 |
+
"logps/chosen": -360.9166564941406,
|
| 253 |
+
"logps/rejected": -359.21478271484375,
|
| 254 |
+
"loss": 0.393,
|
| 255 |
+
"rewards/accuracies": 0.9375,
|
| 256 |
+
"rewards/chosen": -0.12530414760112762,
|
| 257 |
+
"rewards/margins": 0.8049014210700989,
|
| 258 |
+
"rewards/margins_max": 1.5037436485290527,
|
| 259 |
+
"rewards/margins_min": 0.22369906306266785,
|
| 260 |
+
"rewards/margins_std": 0.5746256709098816,
|
| 261 |
+
"rewards/rejected": -0.9302055239677429,
|
| 262 |
+
"step": 120
|
| 263 |
+
},
|
| 264 |
+
{
|
| 265 |
+
"epoch": 0.37,
|
| 266 |
+
"grad_norm": 2.676765116222706,
|
| 267 |
+
"learning_rate": 4.0031170782990214e-06,
|
| 268 |
+
"logits/chosen": -2.488229751586914,
|
| 269 |
+
"logits/rejected": -2.4302055835723877,
|
| 270 |
+
"logps/chosen": -412.65087890625,
|
| 271 |
+
"logps/rejected": -429.1094665527344,
|
| 272 |
+
"loss": 0.3581,
|
| 273 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 274 |
+
"rewards/chosen": -0.3028665781021118,
|
| 275 |
+
"rewards/margins": 0.9839191436767578,
|
| 276 |
+
"rewards/margins_max": 1.651447057723999,
|
| 277 |
+
"rewards/margins_min": 0.29168885946273804,
|
| 278 |
+
"rewards/margins_std": 0.603247344493866,
|
| 279 |
+
"rewards/rejected": -1.2867857217788696,
|
| 280 |
+
"step": 130
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.39,
|
| 284 |
+
"grad_norm": 8.271231871514864,
|
| 285 |
+
"learning_rate": 3.7994171571810756e-06,
|
| 286 |
+
"logits/chosen": -2.3792262077331543,
|
| 287 |
+
"logits/rejected": -2.348208427429199,
|
| 288 |
+
"logps/chosen": -337.531005859375,
|
| 289 |
+
"logps/rejected": -398.3355407714844,
|
| 290 |
+
"loss": 0.3623,
|
| 291 |
+
"rewards/accuracies": 0.949999988079071,
|
| 292 |
+
"rewards/chosen": -0.19116966426372528,
|
| 293 |
+
"rewards/margins": 1.0936906337738037,
|
| 294 |
+
"rewards/margins_max": 1.9557828903198242,
|
| 295 |
+
"rewards/margins_min": 0.2191341668367386,
|
| 296 |
+
"rewards/margins_std": 0.7831128835678101,
|
| 297 |
+
"rewards/rejected": -1.284860372543335,
|
| 298 |
+
"step": 140
|
| 299 |
+
},
|
| 300 |
+
{
|
| 301 |
+
"epoch": 0.42,
|
| 302 |
+
"grad_norm": 6.267090889931009,
|
| 303 |
+
"learning_rate": 3.5831246207606597e-06,
|
| 304 |
+
"logits/chosen": -2.244743585586548,
|
| 305 |
+
"logits/rejected": -2.207181453704834,
|
| 306 |
+
"logps/chosen": -340.91412353515625,
|
| 307 |
+
"logps/rejected": -383.2621765136719,
|
| 308 |
+
"loss": 0.3039,
|
| 309 |
+
"rewards/accuracies": 0.949999988079071,
|
| 310 |
+
"rewards/chosen": -0.5222915410995483,
|
| 311 |
+
"rewards/margins": 1.1795024871826172,
|
| 312 |
+
"rewards/margins_max": 2.02280855178833,
|
| 313 |
+
"rewards/margins_min": 0.273987352848053,
|
| 314 |
+
"rewards/margins_std": 0.7898725271224976,
|
| 315 |
+
"rewards/rejected": -1.701793909072876,
|
| 316 |
+
"step": 150
|
| 317 |
+
},
|
| 318 |
+
{
|
| 319 |
+
"epoch": 0.45,
|
| 320 |
+
"grad_norm": 5.800242307139694,
|
| 321 |
+
"learning_rate": 3.3563355539546795e-06,
|
| 322 |
+
"logits/chosen": -2.081348419189453,
|
| 323 |
+
"logits/rejected": -2.017423629760742,
|
| 324 |
+
"logps/chosen": -342.2528991699219,
|
| 325 |
+
"logps/rejected": -432.951171875,
|
| 326 |
+
"loss": 0.2389,
|
| 327 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 328 |
+
"rewards/chosen": -0.41706037521362305,
|
| 329 |
+
"rewards/margins": 1.5685182809829712,
|
| 330 |
+
"rewards/margins_max": 2.661461114883423,
|
| 331 |
+
"rewards/margins_min": 0.5422362685203552,
|
| 332 |
+
"rewards/margins_std": 0.971255898475647,
|
| 333 |
+
"rewards/rejected": -1.9855785369873047,
|
| 334 |
+
"step": 160
|
| 335 |
+
},
|
| 336 |
+
{
|
| 337 |
+
"epoch": 0.48,
|
| 338 |
+
"grad_norm": 10.002190159330027,
|
| 339 |
+
"learning_rate": 3.121247763262235e-06,
|
| 340 |
+
"logits/chosen": -2.031193971633911,
|
| 341 |
+
"logits/rejected": -1.9694950580596924,
|
| 342 |
+
"logps/chosen": -359.4656982421875,
|
| 343 |
+
"logps/rejected": -506.41217041015625,
|
| 344 |
+
"loss": 0.2274,
|
| 345 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 346 |
+
"rewards/chosen": -0.2984406650066376,
|
| 347 |
+
"rewards/margins": 1.8030246496200562,
|
| 348 |
+
"rewards/margins_max": 2.6862552165985107,
|
| 349 |
+
"rewards/margins_min": 0.6847059726715088,
|
| 350 |
+
"rewards/margins_std": 0.9101365208625793,
|
| 351 |
+
"rewards/rejected": -2.1014654636383057,
|
| 352 |
+
"step": 170
|
| 353 |
+
},
|
| 354 |
+
{
|
| 355 |
+
"epoch": 0.51,
|
| 356 |
+
"grad_norm": 7.495552270047072,
|
| 357 |
+
"learning_rate": 2.8801394778833475e-06,
|
| 358 |
+
"logits/chosen": -1.804452657699585,
|
| 359 |
+
"logits/rejected": -1.6888850927352905,
|
| 360 |
+
"logps/chosen": -424.28533935546875,
|
| 361 |
+
"logps/rejected": -582.659423828125,
|
| 362 |
+
"loss": 0.2208,
|
| 363 |
+
"rewards/accuracies": 1.0,
|
| 364 |
+
"rewards/chosen": -0.923845112323761,
|
| 365 |
+
"rewards/margins": 1.9609006643295288,
|
| 366 |
+
"rewards/margins_max": 3.038311719894409,
|
| 367 |
+
"rewards/margins_min": 0.9176443219184875,
|
| 368 |
+
"rewards/margins_std": 0.9313791394233704,
|
| 369 |
+
"rewards/rejected": -2.8847455978393555,
|
| 370 |
+
"step": 180
|
| 371 |
+
},
|
| 372 |
+
{
|
| 373 |
+
"epoch": 0.54,
|
| 374 |
+
"grad_norm": 5.26489542539375,
|
| 375 |
+
"learning_rate": 2.6353472714635443e-06,
|
| 376 |
+
"logits/chosen": -1.6563403606414795,
|
| 377 |
+
"logits/rejected": -1.4980707168579102,
|
| 378 |
+
"logps/chosen": -365.2936096191406,
|
| 379 |
+
"logps/rejected": -520.5008544921875,
|
| 380 |
+
"loss": 0.1887,
|
| 381 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 382 |
+
"rewards/chosen": -0.4333272874355316,
|
| 383 |
+
"rewards/margins": 2.3968050479888916,
|
| 384 |
+
"rewards/margins_max": 3.5410873889923096,
|
| 385 |
+
"rewards/margins_min": 1.2592780590057373,
|
| 386 |
+
"rewards/margins_std": 1.0371885299682617,
|
| 387 |
+
"rewards/rejected": -2.830132007598877,
|
| 388 |
+
"step": 190
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 0.56,
|
| 392 |
+
"grad_norm": 6.247380737778327,
|
| 393 |
+
"learning_rate": 2.3892434184240536e-06,
|
| 394 |
+
"logits/chosen": -1.6041126251220703,
|
| 395 |
+
"logits/rejected": -1.4733774662017822,
|
| 396 |
+
"logps/chosen": -478.11541748046875,
|
| 397 |
+
"logps/rejected": -663.8895874023438,
|
| 398 |
+
"loss": 0.1872,
|
| 399 |
+
"rewards/accuracies": 0.925000011920929,
|
| 400 |
+
"rewards/chosen": -1.3858261108398438,
|
| 401 |
+
"rewards/margins": 2.5970957279205322,
|
| 402 |
+
"rewards/margins_max": 4.095944404602051,
|
| 403 |
+
"rewards/margins_min": 0.5976935625076294,
|
| 404 |
+
"rewards/margins_std": 1.6118961572647095,
|
| 405 |
+
"rewards/rejected": -3.982921600341797,
|
| 406 |
+
"step": 200
|
| 407 |
+
},
|
| 408 |
+
{
|
| 409 |
+
"epoch": 0.56,
|
| 410 |
+
"eval_logits/chosen": -1.4771034717559814,
|
| 411 |
+
"eval_logits/rejected": -1.4158743619918823,
|
| 412 |
+
"eval_logps/chosen": -568.5177612304688,
|
| 413 |
+
"eval_logps/rejected": -612.582763671875,
|
| 414 |
+
"eval_loss": 0.7859476804733276,
|
| 415 |
+
"eval_rewards/accuracies": 0.6629999876022339,
|
| 416 |
+
"eval_rewards/chosen": -2.839242935180664,
|
| 417 |
+
"eval_rewards/margins": 0.70079505443573,
|
| 418 |
+
"eval_rewards/margins_max": 3.4374256134033203,
|
| 419 |
+
"eval_rewards/margins_min": -2.182706356048584,
|
| 420 |
+
"eval_rewards/margins_std": 1.9158276319503784,
|
| 421 |
+
"eval_rewards/rejected": -3.5400378704071045,
|
| 422 |
+
"eval_runtime": 428.5233,
|
| 423 |
+
"eval_samples_per_second": 4.667,
|
| 424 |
+
"eval_steps_per_second": 0.292,
|
| 425 |
+
"step": 200
|
| 426 |
+
},
|
| 427 |
+
{
|
| 428 |
+
"epoch": 0.59,
|
| 429 |
+
"grad_norm": 5.190377965258524,
|
| 430 |
+
"learning_rate": 2.1442129043167877e-06,
|
| 431 |
+
"logits/chosen": -1.4293570518493652,
|
| 432 |
+
"logits/rejected": -1.3583998680114746,
|
| 433 |
+
"logps/chosen": -448.363037109375,
|
| 434 |
+
"logps/rejected": -679.0516357421875,
|
| 435 |
+
"loss": 0.1699,
|
| 436 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 437 |
+
"rewards/chosen": -1.281787395477295,
|
| 438 |
+
"rewards/margins": 2.944481134414673,
|
| 439 |
+
"rewards/margins_max": 4.665500640869141,
|
| 440 |
+
"rewards/margins_min": 0.9343290328979492,
|
| 441 |
+
"rewards/margins_std": 1.711601972579956,
|
| 442 |
+
"rewards/rejected": -4.226268291473389,
|
| 443 |
+
"step": 210
|
| 444 |
+
},
|
| 445 |
+
{
|
| 446 |
+
"epoch": 0.62,
|
| 447 |
+
"grad_norm": 4.464327030535654,
|
| 448 |
+
"learning_rate": 1.9026303129961049e-06,
|
| 449 |
+
"logits/chosen": -1.474830150604248,
|
| 450 |
+
"logits/rejected": -1.3183205127716064,
|
| 451 |
+
"logps/chosen": -591.7376708984375,
|
| 452 |
+
"logps/rejected": -873.0149536132812,
|
| 453 |
+
"loss": 0.1359,
|
| 454 |
+
"rewards/accuracies": 0.949999988079071,
|
| 455 |
+
"rewards/chosen": -2.382814407348633,
|
| 456 |
+
"rewards/margins": 3.604060411453247,
|
| 457 |
+
"rewards/margins_max": 4.968972206115723,
|
| 458 |
+
"rewards/margins_min": 1.4786490201950073,
|
| 459 |
+
"rewards/margins_std": 1.6280790567398071,
|
| 460 |
+
"rewards/rejected": -5.986874580383301,
|
| 461 |
+
"step": 220
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.65,
|
| 465 |
+
"grad_norm": 11.671302383042187,
|
| 466 |
+
"learning_rate": 1.66683681459314e-06,
|
| 467 |
+
"logits/chosen": -1.4693548679351807,
|
| 468 |
+
"logits/rejected": -1.3056604862213135,
|
| 469 |
+
"logps/chosen": -644.6790771484375,
|
| 470 |
+
"logps/rejected": -878.6692504882812,
|
| 471 |
+
"loss": 0.146,
|
| 472 |
+
"rewards/accuracies": 0.9375,
|
| 473 |
+
"rewards/chosen": -2.730034351348877,
|
| 474 |
+
"rewards/margins": 3.4357237815856934,
|
| 475 |
+
"rewards/margins_max": 5.224689483642578,
|
| 476 |
+
"rewards/margins_min": 1.111033320426941,
|
| 477 |
+
"rewards/margins_std": 1.8828909397125244,
|
| 478 |
+
"rewards/rejected": -6.165757656097412,
|
| 479 |
+
"step": 230
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 0.68,
|
| 483 |
+
"grad_norm": 5.647778822191043,
|
| 484 |
+
"learning_rate": 1.4391174773015836e-06,
|
| 485 |
+
"logits/chosen": -1.4095017910003662,
|
| 486 |
+
"logits/rejected": -1.2933498620986938,
|
| 487 |
+
"logps/chosen": -636.0205688476562,
|
| 488 |
+
"logps/rejected": -929.3019409179688,
|
| 489 |
+
"loss": 0.1482,
|
| 490 |
+
"rewards/accuracies": 0.949999988079071,
|
| 491 |
+
"rewards/chosen": -3.1122348308563232,
|
| 492 |
+
"rewards/margins": 3.3490359783172607,
|
| 493 |
+
"rewards/margins_max": 5.252045631408691,
|
| 494 |
+
"rewards/margins_min": 1.3964684009552002,
|
| 495 |
+
"rewards/margins_std": 1.7266426086425781,
|
| 496 |
+
"rewards/rejected": -6.461270809173584,
|
| 497 |
+
"step": 240
|
| 498 |
+
},
|
| 499 |
+
{
|
| 500 |
+
"epoch": 0.7,
|
| 501 |
+
"grad_norm": 29.012737765276803,
|
| 502 |
+
"learning_rate": 1.2216791228457778e-06,
|
| 503 |
+
"logits/chosen": -1.3994617462158203,
|
| 504 |
+
"logits/rejected": -1.2709558010101318,
|
| 505 |
+
"logps/chosen": -621.2999267578125,
|
| 506 |
+
"logps/rejected": -928.4699096679688,
|
| 507 |
+
"loss": 0.1235,
|
| 508 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 509 |
+
"rewards/chosen": -3.1001884937286377,
|
| 510 |
+
"rewards/margins": 3.6740684509277344,
|
| 511 |
+
"rewards/margins_max": 5.534354209899902,
|
| 512 |
+
"rewards/margins_min": 1.508026361465454,
|
| 513 |
+
"rewards/margins_std": 1.7670818567276,
|
| 514 |
+
"rewards/rejected": -6.774256706237793,
|
| 515 |
+
"step": 250
|
| 516 |
+
},
|
| 517 |
+
{
|
| 518 |
+
"epoch": 0.73,
|
| 519 |
+
"grad_norm": 4.553216346552263,
|
| 520 |
+
"learning_rate": 1.0166289402331391e-06,
|
| 521 |
+
"logits/chosen": -1.4789546728134155,
|
| 522 |
+
"logits/rejected": -1.3316059112548828,
|
| 523 |
+
"logps/chosen": -634.4136962890625,
|
| 524 |
+
"logps/rejected": -945.2862548828125,
|
| 525 |
+
"loss": 0.1438,
|
| 526 |
+
"rewards/accuracies": 0.9375,
|
| 527 |
+
"rewards/chosen": -3.4232017993927,
|
| 528 |
+
"rewards/margins": 3.440350294113159,
|
| 529 |
+
"rewards/margins_max": 5.2788987159729,
|
| 530 |
+
"rewards/margins_min": 1.4436607360839844,
|
| 531 |
+
"rewards/margins_std": 1.7658751010894775,
|
| 532 |
+
"rewards/rejected": -6.863552093505859,
|
| 533 |
+
"step": 260
|
| 534 |
+
},
|
| 535 |
+
{
|
| 536 |
+
"epoch": 0.76,
|
| 537 |
+
"grad_norm": 6.13292881557691,
|
| 538 |
+
"learning_rate": 8.259540650444736e-07,
|
| 539 |
+
"logits/chosen": -1.3842976093292236,
|
| 540 |
+
"logits/rejected": -1.286069631576538,
|
| 541 |
+
"logps/chosen": -662.1298217773438,
|
| 542 |
+
"logps/rejected": -982.2066650390625,
|
| 543 |
+
"loss": 0.1621,
|
| 544 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 545 |
+
"rewards/chosen": -3.530142307281494,
|
| 546 |
+
"rewards/margins": 3.6669273376464844,
|
| 547 |
+
"rewards/margins_max": 5.465144157409668,
|
| 548 |
+
"rewards/margins_min": 1.4210524559020996,
|
| 549 |
+
"rewards/margins_std": 1.8017559051513672,
|
| 550 |
+
"rewards/rejected": -7.197070121765137,
|
| 551 |
+
"step": 270
|
| 552 |
+
},
|
| 553 |
+
{
|
| 554 |
+
"epoch": 0.79,
|
| 555 |
+
"grad_norm": 8.782327714810695,
|
| 556 |
+
"learning_rate": 6.515023221586722e-07,
|
| 557 |
+
"logits/chosen": -1.3697433471679688,
|
| 558 |
+
"logits/rejected": -1.28987455368042,
|
| 559 |
+
"logps/chosen": -655.1025390625,
|
| 560 |
+
"logps/rejected": -1005.9417724609375,
|
| 561 |
+
"loss": 0.1048,
|
| 562 |
+
"rewards/accuracies": 0.987500011920929,
|
| 563 |
+
"rewards/chosen": -3.516455888748169,
|
| 564 |
+
"rewards/margins": 3.8169894218444824,
|
| 565 |
+
"rewards/margins_max": 5.726956367492676,
|
| 566 |
+
"rewards/margins_min": 1.5342118740081787,
|
| 567 |
+
"rewards/margins_std": 1.8832015991210938,
|
| 568 |
+
"rewards/rejected": -7.3334455490112305,
|
| 569 |
+
"step": 280
|
| 570 |
+
},
|
| 571 |
+
{
|
| 572 |
+
"epoch": 0.82,
|
| 573 |
+
"grad_norm": 16.94834231932571,
|
| 574 |
+
"learning_rate": 4.949643185335288e-07,
|
| 575 |
+
"logits/chosen": -1.4026738405227661,
|
| 576 |
+
"logits/rejected": -1.3095262050628662,
|
| 577 |
+
"logps/chosen": -593.7080078125,
|
| 578 |
+
"logps/rejected": -906.31298828125,
|
| 579 |
+
"loss": 0.1582,
|
| 580 |
+
"rewards/accuracies": 0.9375,
|
| 581 |
+
"rewards/chosen": -3.067917823791504,
|
| 582 |
+
"rewards/margins": 3.355029344558716,
|
| 583 |
+
"rewards/margins_max": 5.543763637542725,
|
| 584 |
+
"rewards/margins_min": 1.00192391872406,
|
| 585 |
+
"rewards/margins_std": 2.0951552391052246,
|
| 586 |
+
"rewards/rejected": -6.422947883605957,
|
| 587 |
+
"step": 290
|
| 588 |
+
},
|
| 589 |
+
{
|
| 590 |
+
"epoch": 0.85,
|
| 591 |
+
"grad_norm": 9.830855192004018,
|
| 592 |
+
"learning_rate": 3.578570595810274e-07,
|
| 593 |
+
"logits/chosen": -1.5070269107818604,
|
| 594 |
+
"logits/rejected": -1.3751003742218018,
|
| 595 |
+
"logps/chosen": -646.901123046875,
|
| 596 |
+
"logps/rejected": -984.0985107421875,
|
| 597 |
+
"loss": 0.1102,
|
| 598 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 599 |
+
"rewards/chosen": -3.0038809776306152,
|
| 600 |
+
"rewards/margins": 3.9427695274353027,
|
| 601 |
+
"rewards/margins_max": 5.859181880950928,
|
| 602 |
+
"rewards/margins_min": 1.6614525318145752,
|
| 603 |
+
"rewards/margins_std": 1.9274225234985352,
|
| 604 |
+
"rewards/rejected": -6.946650505065918,
|
| 605 |
+
"step": 300
|
| 606 |
+
},
|
| 607 |
+
{
|
| 608 |
+
"epoch": 0.85,
|
| 609 |
+
"eval_logits/chosen": -1.4022948741912842,
|
| 610 |
+
"eval_logits/rejected": -1.3416800498962402,
|
| 611 |
+
"eval_logps/chosen": -773.4503173828125,
|
| 612 |
+
"eval_logps/rejected": -840.8230590820312,
|
| 613 |
+
"eval_loss": 0.8934583067893982,
|
| 614 |
+
"eval_rewards/accuracies": 0.6470000147819519,
|
| 615 |
+
"eval_rewards/chosen": -4.888567924499512,
|
| 616 |
+
"eval_rewards/margins": 0.9338728189468384,
|
| 617 |
+
"eval_rewards/margins_max": 4.508862495422363,
|
| 618 |
+
"eval_rewards/margins_min": -2.803253412246704,
|
| 619 |
+
"eval_rewards/margins_std": 2.4985644817352295,
|
| 620 |
+
"eval_rewards/rejected": -5.822441577911377,
|
| 621 |
+
"eval_runtime": 428.3228,
|
| 622 |
+
"eval_samples_per_second": 4.669,
|
| 623 |
+
"eval_steps_per_second": 0.292,
|
| 624 |
+
"step": 300
|
| 625 |
+
},
|
| 626 |
+
{
|
| 627 |
+
"epoch": 0.87,
|
| 628 |
+
"grad_norm": 25.66825260695654,
|
| 629 |
+
"learning_rate": 2.4150924791035037e-07,
|
| 630 |
+
"logits/chosen": -1.4382137060165405,
|
| 631 |
+
"logits/rejected": -1.2731282711029053,
|
| 632 |
+
"logps/chosen": -688.03857421875,
|
| 633 |
+
"logps/rejected": -948.1687622070312,
|
| 634 |
+
"loss": 0.1452,
|
| 635 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 636 |
+
"rewards/chosen": -3.9022278785705566,
|
| 637 |
+
"rewards/margins": 3.437220335006714,
|
| 638 |
+
"rewards/margins_max": 5.5707316398620605,
|
| 639 |
+
"rewards/margins_min": 1.006291389465332,
|
| 640 |
+
"rewards/margins_std": 2.0285696983337402,
|
| 641 |
+
"rewards/rejected": -7.339447975158691,
|
| 642 |
+
"step": 310
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.9,
|
| 646 |
+
"grad_norm": 24.605253871204138,
|
| 647 |
+
"learning_rate": 1.4704840690808658e-07,
|
| 648 |
+
"logits/chosen": -1.4029080867767334,
|
| 649 |
+
"logits/rejected": -1.3075344562530518,
|
| 650 |
+
"logps/chosen": -682.196044921875,
|
| 651 |
+
"logps/rejected": -989.9052734375,
|
| 652 |
+
"loss": 0.1673,
|
| 653 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 654 |
+
"rewards/chosen": -3.739229202270508,
|
| 655 |
+
"rewards/margins": 3.553121566772461,
|
| 656 |
+
"rewards/margins_max": 5.583710670471191,
|
| 657 |
+
"rewards/margins_min": 0.8181284070014954,
|
| 658 |
+
"rewards/margins_std": 2.200822114944458,
|
| 659 |
+
"rewards/rejected": -7.292350769042969,
|
| 660 |
+
"step": 320
|
| 661 |
+
},
|
| 662 |
+
{
|
| 663 |
+
"epoch": 0.93,
|
| 664 |
+
"grad_norm": 9.663524467768484,
|
| 665 |
+
"learning_rate": 7.538995394063996e-08,
|
| 666 |
+
"logits/chosen": -1.5248098373413086,
|
| 667 |
+
"logits/rejected": -1.3775807619094849,
|
| 668 |
+
"logps/chosen": -683.1748046875,
|
| 669 |
+
"logps/rejected": -983.3424682617188,
|
| 670 |
+
"loss": 0.1405,
|
| 671 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 672 |
+
"rewards/chosen": -3.2885494232177734,
|
| 673 |
+
"rewards/margins": 3.875479221343994,
|
| 674 |
+
"rewards/margins_max": 5.90716552734375,
|
| 675 |
+
"rewards/margins_min": 1.6211687326431274,
|
| 676 |
+
"rewards/margins_std": 1.9896876811981201,
|
| 677 |
+
"rewards/rejected": -7.164028167724609,
|
| 678 |
+
"step": 330
|
| 679 |
+
},
|
| 680 |
+
{
|
| 681 |
+
"epoch": 0.96,
|
| 682 |
+
"grad_norm": 3.7848904322458856,
|
| 683 |
+
"learning_rate": 2.722832907015971e-08,
|
| 684 |
+
"logits/chosen": -1.389829397201538,
|
| 685 |
+
"logits/rejected": -1.2816789150238037,
|
| 686 |
+
"logps/chosen": -664.4754638671875,
|
| 687 |
+
"logps/rejected": -1003.7318115234375,
|
| 688 |
+
"loss": 0.115,
|
| 689 |
+
"rewards/accuracies": 0.987500011920929,
|
| 690 |
+
"rewards/chosen": -3.6462998390197754,
|
| 691 |
+
"rewards/margins": 3.879322052001953,
|
| 692 |
+
"rewards/margins_max": 5.659389019012451,
|
| 693 |
+
"rewards/margins_min": 1.880358338356018,
|
| 694 |
+
"rewards/margins_std": 1.7462043762207031,
|
| 695 |
+
"rewards/rejected": -7.5256218910217285,
|
| 696 |
+
"step": 340
|
| 697 |
+
},
|
| 698 |
+
{
|
| 699 |
+
"epoch": 0.99,
|
| 700 |
+
"grad_norm": 5.242067065610791,
|
| 701 |
+
"learning_rate": 3.030265255329623e-09,
|
| 702 |
+
"logits/chosen": -1.377044439315796,
|
| 703 |
+
"logits/rejected": -1.2859034538269043,
|
| 704 |
+
"logps/chosen": -652.96337890625,
|
| 705 |
+
"logps/rejected": -1036.343505859375,
|
| 706 |
+
"loss": 0.1061,
|
| 707 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 708 |
+
"rewards/chosen": -3.380042552947998,
|
| 709 |
+
"rewards/margins": 4.118873119354248,
|
| 710 |
+
"rewards/margins_max": 5.629302501678467,
|
| 711 |
+
"rewards/margins_min": 2.122800827026367,
|
| 712 |
+
"rewards/margins_std": 1.6110029220581055,
|
| 713 |
+
"rewards/rejected": -7.4989166259765625,
|
| 714 |
+
"step": 350
|
| 715 |
+
},
|
| 716 |
+
{
|
| 717 |
+
"epoch": 1.0,
|
| 718 |
+
"step": 355,
|
| 719 |
+
"total_flos": 0.0,
|
| 720 |
+
"train_loss": 0.3073710792501208,
|
| 721 |
+
"train_runtime": 4217.5244,
|
| 722 |
+
"train_samples_per_second": 1.346,
|
| 723 |
+
"train_steps_per_second": 0.084
|
| 724 |
+
}
|
| 725 |
+
],
|
| 726 |
+
"logging_steps": 10,
|
| 727 |
+
"max_steps": 355,
|
| 728 |
+
"num_input_tokens_seen": 0,
|
| 729 |
+
"num_train_epochs": 1,
|
| 730 |
+
"save_steps": 100,
|
| 731 |
+
"total_flos": 0.0,
|
| 732 |
+
"train_batch_size": 4,
|
| 733 |
+
"trial_name": null,
|
| 734 |
+
"trial_params": null
|
| 735 |
+
}
|